What is in the Schmidt & Hunter 1998 validity table, and is it still current?
A 1998 meta-analysis that tabulated the predictive validity of nineteen hiring methods and the gain from combining each with a general mental ability test. Its figures — cognitive ability at .51, work samples at .54, structured interviews at .51 — became the reference table for personnel selection and were substantially revised downward in 2022.
Schmidt, F. L. & Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124(2), 262–274.
Primary source opened and quotes confirmed on .
In its own words
“The value of .51 in Table 1 for the validity of GMA is from a very large meta-analytic study conducted for the U.S. Department of Labor.”
What it does not say
Each of these is a claim made in this market and attributed to the source above. None of them is supported by it.
Commonly claimed: That .51 is the current best estimate for cognitive ability testing.
It was the best estimate for twenty-four years. Sackett, Zhang, Berry and Lievens (2022) put the same predictor at .31 after correcting a systematic overcorrection for range restriction. A 2026 page quoting .51 without saying it has been revised is quoting a superseded number.
Commonly claimed: That the second column shows what a test adds to your process.
The multiple-R column shows what each method adds to a cognitive ability test, not what it adds to the interview or CV screen an employer is already running. It answers a different question from the one most buyers have.
Commonly claimed: That two instruments with similar numbers are interchangeable.
Read the two columns together and they are not. Integrity tests carry a modest standalone .41 and the largest combined figure in the table at .65, because they measure something close to uncorrelated with ability. The buying rule is to add the predictor least like what you already have.
The figures
Table 1, as printed: validity for predicting job performance, and multiple R when combined with a GMA test
| Method | Validity (r) | Multiple R with GMA |
|---|---|---|
| Work sample tests | .54 | .63 |
| GMA tests | .51 | — |
| Structured interviews | .51 | .63 |
| Peer ratings | .49 | .58 |
| Job knowledge tests | .48 | .58 |
| Job tryout procedure | .44 | .58 |
| Integrity tests | .41 | .65 |
| Unstructured interviews | .38 | .55 |
| Assessment centres | .37 | .53 |
| Biographical data | .35 | .52 |
| Conscientiousness tests | .31 | .60 |
| Reference checks | .26 | .57 |
| Job experience (years) | .18 | .54 |
| Years of education | .10 | .52 |
| Interests | .10 | .51 |
| Graphology | .02 | .51 |
| Age | −.01 | .51 |
Read from the full text of the paper on 16 September 2026. Rows are reproduced for reference; the values are the 1998 estimates and are not the field's current ones.
Why this table is worth indexing rather than quoting
It is the single most cited artefact in the hiring-assessment market, and the version that circulates is almost always a partial one. Vendor pages reproduce the rows that flatter the instrument being sold and omit the combination column that shows an entirely different instrument adding more.
Publishing the table intact, with the date it was superseded attached, is more useful than either quoting it or ignoring it. A buyer who has seen a figure from it in a sales deck can come here, find the rest of the row, and ask the two questions that follow: over what, and as of when.
What survived the 2022 revision
More than the headlines suggest. The 2022 re-analysis states that most procedures ranking high in earlier summaries remain high in rank. The structural insights of the 1998 paper — that predictors combine better when they are uncorrelated, that unstructured interviews underperform structured ones, that graphology predicts nothing — are not in dispute.
What did not survive is the level of every coefficient and the framing that made cognitive ability the anchor other methods add to. Treat the 1998 paper as the source of the argument and the 2022 paper as the source of the numbers.
Where this source is used here
These pages argue from the source above. If it is ever superseded, these are the pages that have to change.
Read next
- Sackett et al. (2022) — the validity re-estimates — The paper that cut the published validity of most hiring predictors by .10–.20 and put structured interviews on top.
- Sackett et al. (2023) — what the revision means for a screen — The follow-up that prices the decision: dropping cognitive ability from a six-predictor composite costs .05 of validity, not .20.
Founder of AssessAll and of Bodhih Training Solutions, a corporate training company in Bangalore. Works on assessment design, scoring and reporting across hiring, L&D and certification programmes.
Last reviewed